Elucidation of the ageing chemistry of whiskey aged in oak barrels is important for an understanding of the final aroma and taste. Both volatile and non-volatile species comprise the additional compounds produced during maturation and their analysis requires the complimentary techniques of gas chromatography mass spectrometry (GC-MS) and liquid chromatography mass spectrometry (LC-MS). The recently introduced technology of GC large volume injection (LVI) with solvent venting is used for direct injection of samples without extensive sample pre-treatment. Data deconvolution of the resulting total ion chromatogram (TIC) allows rapid isolation of targets from other co-eluting - mainly fermentation - compounds. LC-MS then reveals additional subtle differences in compound profiles in whiskey matured in oak barrels from different wood species. Ion trap (IT) and time-of-flight (TOF) mass spectrometry (MS) techniques can than be used to give structural information and empirical formula determination on these relatively unknown species. A final important tool is use of an LC active splitter preparative system, where the specific MS signal of a compound is the trigger for isolation and collection. In this way unknown compounds can be collected in high purity for both further spectroscopic analysis and sensory evaluation.

For a premium commercial spirit, such as whiskey, wood ageing or maturation in oak barrels for extended periods is essential to give it the unique taste that consumers like.The various processing stages of barley brewing, fermentation and distillation all contribute a unique sub-set of flavours to the spirit, but it is clear that the final long ageing step in oak barrels has a significant influence on the final distinctive aroma and taste. During maturation the flavour compound changes that contribute to taste (1,2) can be subdivided into four areas:   

The reaction of wood and wood compounds with components of the unaged distillate, including the dominant component: ethanol

Reactions involving only wood extractives

The processes involved and the final outcome are made more complex by additional variables, including the pre-treatment of the barrel when new, the type and origin of the wood used, and whether the barrel has been previously used in a maturation cycle (3) 

Pre-treatment of the barrel involves heating the inside of the cask and may vary from an intense charring over a fire to more gentle toasting with hot air. The intensity of heating will affect the type and levels of compounds produced from the degradation of wood lignin and macromolecules. In addition to heat pre-treatment an additional mechanism that contributes to flavour involves the production of a complex from wood lignin and ethanol in which ethanol acts as a solvent and a reactant. 

The mild acidic hydrolysis of this complex produces simple aromatic phenolic compounds. The majority of published work in this area to date has concentrated on elucidating the mechanisms involved in this degradation of wood to give simpler flavour molecules. An examination of the literature indicates many studies on the development, determination, structural elucidation and sensory properties of oak wood ellagitannins, a large group of polyphenolic compounds widely distributed in plants (4,5). The usual practice is to chemically extract these ellagitannins directly from the corresponding heartwood of the oak species that has initially been converted into oak chip or sawdust form. Interesting results are presented but practically no linkage is established between the species identified and their possible presence in spirits that have undergone a "natural" maturation cycle. Methanol-water extracts of oak wood chips, and wine and spirit samples actually aged with oak chips have also been used to analyse and determine low-molecular-weight phenolic compounds (6,7). Again, the problem is that this much shorter, forced ageing with a hugely increased wood-to-spirit surface exposure cannot replicate the intricate complexity and result of a natural ageing cycle. 

In this article we concentrate entirely on the comparative analysis of the actual pot distillate at various intervals of barrel maturation extending to 20 years in both new 

 (European) wood. Six of each of these barrels were made to identical dimensions of 500 litres volume, were lightly toasted on the inside and had no additional treatments or seasoning to uniquely allow the wood effect to be expressed in the subsequent analyses.  

 Figure 1: Visual changes in unaged parent whiskey (centre) after 20 years in new Quercus robur (left) and Quercus alba oak barrels. 

Figure 1 visually compares the unaged spirit (parent) before cask filling with the equivalent at 20 years from the two wood species and clearly illustrates the remarkable influence of the wood. Needless to say the sensory properties of the two aged samples in terms of both aroma and taste are also dramatically different. Gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) analysis were performed. GC-MS was specifically used to investigate compounds associated with wood toasting and lignin breakdown. In our opinion comparative analysis of those compounds, mainly of fermentation origin, originally present in the unaged distillate, provides only limited information in the context of changes induced by maturation.  

Direct large volume injection (LVI) of actual spirit samples with solvent venting of the ethanol-water matrix was used for GC-MS analysis. This injection technology removes the need for extensive sample preparation and concentration, and offers a boiling point recovery of target analytes when correctly optimized. Less volatile compounds were analysed by different LC-MS techniques using atmospheric pressure chemical ionization (APCI) in the negative mode. Ion trap mass spectrometry (IT-MS-MS) was used for structural information and the accurate mass time-of-flight (TOF) technique was used to determine the compound empirical formula. 

The most widely applied LVI technique uses a PTV (programmable temperature vaporization) inlet in solvent vent mode and important parameters are the PTV temperature, the carrier gas venting flow and the sample introduction rate. If the solvent is not removed at the same time as a vapour as it enters the injector liner, then analytes will simply be removed with liquid solvent through the split line. The goal is to recover and retain the maximum amount of analytes from the solvent for subsequent splitless transfer to the column.  

Research groups in Germany and Holland have pioneered research into the fundamentals of this technique and have explained the interactions between the variables involved (8-10). In the case of distilled spirits the "solvent" will be the natural ethanol-water matrix of the sample and while the maximum rate of injection for a particular solvent can be theoretically calculated from the solvent vapour pressure, split gas venting flow and the PTV temperature, an additional reduction of this maximum is necessary when water is present. This is because the high latent heat of vaporization of water induces an additional cooling effect in the liner, and the venting gas flow will have a lower than theoretical saturation with water vapour.  

For the results described in this article 100 µL of spirit samples were successfully injected with 60% water at a PTV temperature of 40 °C, a solvent vent gas flow of 500 mL/min and an injection rate of 12 µL/min. This latter parameter is critical because doubling the injection speed results in more than 80% of the analyte being lost compared with the optimum injection rate of 12 µL/min. Under optimum conditions recovery data for 100 µL injections of a compound test mix compared with standard 1 µL injections of prorata concentrated solutions show more than 90% recovery of analytes and good area reproducibility vis-à-is added internal standards (ISs). Proper column choice is also important and we use an apolar CP-Sil 5 CB (Agilent Technologies, Santa Clara, California, USA) with dimensions 25 m × 0.15 mm i.d. × 2.0 µm d

Table 1: Database for target compound GCâMS deconvolution after large volume injection with solvent venting.  

Injection of 100 µL will also transfer substantial amounts of non-target ester and acid analytes to the column and the thick film with phase ratio β = 19 provides the high capacity needed for retention of good analyte peak shape. The large injections were performed on disposable 2 mm i.d. plain glass liners packed with glass wool. This is useful as more frequent liner changes are obviously necessary. Unattended sequences can be accommodated by using an automated liner exchange module (Gerstel, Mülheim an der Ruhr, Germany). 

A recent publication (11) describes a 5–10 µL solvent injection on an acidic wax column of distilled spirits for major fermentation esters and acids and a limited number of minor compounds associated with cereal raw material preparation and wood ageing. The present work is an extension of this successful approach, but introducing much larger volumes requiring injection speed programming and using a high temperature apolar phase for elution of less volatile analytes. Finally, because the mass spectra of target analytes will be known, expected fragment ion ratios can be used by deconvolution software (Ion Signature Technology, North Smithfield, Rhode Island, USA) to return a reconstructed ion chromatogram (RIC) of target analytes only, from the generic TIC. In this way a useful synergy is achieved with LVI simplifying the first rate limiting step of sample preparation and deconvolution the second rate limiting step of data reduction. 

Table 1 describes the target compound data base for deconvolution of these compounds from the standard total ion chromatogram (TIC). This database is first established, either from the spectral data of expected compounds (vanillin, etc), literature data on monomeric lignin-derived products after thermal degradation of wood (12,13), or similar compounds after pyrolysis GC–MS of plant and lignin materials (14,15). Lignins may be pyrolyzed reproducibly to produce a mixture of relatively simple phenols resulting from cleavage of ether and certain C-C linkages. Other sources for identification of candidate target compounds are LC–GC–MS and GC–GC–MS (16). Deconvolution processes the scan data file looking for the expected relative abundances (Table 1) associated with each target and provides a list of compounds found and their relative concentrations. This process is more comprehensive than selected ion monitoring (SIM) because the software identifies and discards contributions to the spectra that have not originated from the target compounds. 

Figure 2: Overlay of GC total ion chromatogram (TIC) and corresponding target compound reconstructed ion chromatogram (RIC) after 100 ÂµL injection of aged whiskey. 

Figure 2 shows an overlay of the TIC and associated target RIC for a 100 µL injection of a composite 20-year-old sample from the 

 casks, and demonstrates the ability of this procedure to extract a more meaningful pattern from the standard TIC. 

Figure 3: Comparison of GC RIC traces for 100Â ÂµL injections of aged whiskey from Quercus robur (left) and Quercus albaoak barrels.  

Figure 3 shows a comparison of the target RIC's for the 20 year old 

-matured samples. Additional complexity is apparent for the 

 sample, but this mainly reflects an increased response for many of the same target compounds. This type of rapid profiling can be very useful for cask quality assessment and will also in authenticity verification. We will now look at how LC–MS technology applied to the less volatile compounds reveals unique and interesting differences between these samples. 

In contrast to GC–based studies, relatively few publications feature HPLC–MS or the characterization of important oak wood compounds in aged distilled spirits (17). The principal application has been high performance liquid chromatography with diode array detection (HPLC–DAD) to quantify vanillin-type phenolic aldehyde structures, as well as their associated acids and coumaric acid side chain derivatives (18). Where LC–MS has been used the main ionization mode has been electrospray ionization (ESI), which requires care to avoid problems of ionization suppression and adduct formation. In this study atmospheric pressure chemical ionization (APCI) in the negative mode was used because this provides highly selective detection for compounds with an acidic ionizable phenolic hydroxyl group. Mass spectra show almost no fragmentation and predominantly give the [M–H]

  ion after proton abstraction with much reduced chromatographic background compared with electrospray mode. APCI and electrospray are relatively weak ionization techniques and therefore the molecular ion [M–H] is usually dominant in the mass spectra of compounds. This facilitates the use of extracted ion chromatograms (EIC) of the molecular ion for quantification. Good separation was achieved on a 250 × 4.6 mm Zorbax SP-Phenyl column (Agilent Technologies) in reversed phase (RP) mode using 5 mM formic acid and methanol as eluents. 

Figure 4: Comparison of TICâMS APCI negative traces for aged whiskey from Quercus robur and Quercus alba oak barrels with the corresponding LCâUV trace for aged whiskey from Quercus robur barrels. 

Figure 4 shows the comparison of the TIC–MS chromatograms obtained during the analysis of aged 

 samples and also the UV trace (λ = 290 nm) for an aged 

 sample. Compounds in the region up to 35 min are the welldocumented vanillin-type structures and are detected in both MS and UV modes. More apolar later eluting species, with typical molecular weights from 400–700 Da, are only amenable to MS detection. The TIC–MS chromatograms of whiskey from both woods at the same age shows that, for the earlier eluting phenolic aldehydes and acids, the concentrations are generally higher from 

. Later eluting compounds with higher molecular masses show substantial interwood concentration differences with significantly higher concentrations from 

The previous section has shown how a single mass analysis in APCI negative mode is well suited for method development, basic characterization of the sample composition and quantification of interesting compounds. However, more advanced MS techniques such as ion trap (IT) and time-of-flight (TOF) offer the possibility of obtaining structural information of interesting compounds. Mass spectral databases for LC–MS techniques are generally not available and therefore structure determination of unknown compounds needs to be performed by complex experiments which involve all available LC–MS techniques.

Table 2: Fragmentation and exact mass data on interesting low volatility compounds after LCâIon Trap and LCâTOF. 

Table 2 summarizes all LC–MS data on most of the compounds detected in whiskey from both wood species. With ion trap, up to MS

  fragmentations were performed to reveal interesting substructures of compounds. The masses of these fragments are listed in Table 2. Without changing the chromatography exact mass information was obtained from TOF measurements on the molecular ion [M–H]. Partial overlap of the signals as well as varying concentration can influence the accuracy of the mass determination. Therefore, the exact masses and calculated formulas were recorded only for those non-overlapping main compounds present at fairly high concentration. These are listed in Table 2. 

The use of these techniques can be illustrated by considering the fragmentation of compound 12 (Table 2), 

 = 419, identified as lyoniresinol, and the earlier eluting compound 11, 

 = 581, identified as glucolyoniresinol. Lyoniresinol (19) with its typical phenolic aldehyde structure has four methoxy groups resulting in the expected sequential losses of 

 = 404, 389, 374 and 359. Figure 5 shows the parent lyoniresinol structure and its MS, MS

  spectra together with the corresponding fragmentation patterns for glucolyoniresinol. In general, compounds which are formed by the addition or condensation of smaller molecules to lyoniresinol should show similar fragmentation and in this respect it can be seen that the MS

  spectra respectively of glucolyoniresinol are identical with the MS

  spectra of lyoniresinol. Compound 11 is a derivative of lyoniresinol and is also significantly more polar. The most probable formula obtained from the TOF measurement and proposed chemical structure of glucolyoniresinol is presented in Table 2. The reaction of phenolic compounds with glucose in plants has been extensively described in the literature (20), an additional new example is compound 5 in Table 2, identified as glucogallic acid. An analogous relationship can be established between compound 25 (

  pattern of compound 31. Compound 25 is again a glucoside of compound 31. Additionally, compound 31 has isomers with significantly lower concentrations and among these compound 28 is the isomer with the highest relative concentration and identical fragmentation to compound 31. 

Figure 5: Comparison of Ion Trap LCâMSâMS fragmentation of lyoniresinol and glucolyoniresinol. 

Fragmentation pattern and exact mass experiments can give information on structural details and the possible formula mass of a compound, but for the final determination of the compound chemical structure it may be necessary to isolate the compound for further spectroscopic examination. 

Figure 6: LCâMS configuration for preparative isolation of compounds. 

Compound isolation also provides pure material for sensory examination and spiking at various levels into spirits at various stages of wood maturation. Previous data has shown that the ultra-violet (UV) response of new unknown high molecular compounds is very low and therefore we use preparative LC–MS for the isolation of compounds. This system is outlined in Figure 6 and is based on the "active splitter" principle where the MS signal for the target compound is used to divert the compound to the fraction collector. The fractions were cut with the MS in selected ion mode and using the typical APCI negative 

 of the target. Progress so far has provided the following three main compounds at high purity: Compound 12 (

 = 501) at the milligramme level. Purified samples will be further investigated for structure determination by 

Various novel GC–MS and LC–MS techniques have been used to investigate quite obvious visual and sensory differences in whiskey matured in oak barrels made from different wood species. Substantial information can be obtained by using approaches that allow direct analysis of samples without extensive sample preparation. In GC–MS this is achieved by on-line concentration in the inlet by LVI with solvent venting of the dominant ethanol-water fraction followed by target compound deconvolution, and in LC–MS by using atmospheric pressure chemical ionization (APCI) as a selective detection mechanism for the compounds of interest. LC–IT-MS provides molecular fragmentation information to assist in structure determination and LC–TOFMS provides exact mass information for empirical formula determination.  

, PhD, is Head of Technical Development at Irish Distillers, an important subsidiary of Groupe Pernod Ricard. He has published extensively on modern GC sample introduction techniques applied to distilled spirits, where much larger volumes than standard can be introduced with matrix removal incorporated into the injection procedure. He represents Ireland on the Technical Committee of CEPS, the European Spirits Producers Association. 

, PhD, is a Senior Scientist at Irish Distillers, Pernod Ricard, and has experience in state-of-the-art separation science and chromatographic techniques. She previously worked in the environmental and pharmaceutical sectors with a specialty in LC–MS and LC–TOF-MS techniques. Her main focus is on the methods development, validation and compound identification. 

, PhD, is the owner and general manager of TeLA GmbH, a contract laboratory in the field of Food Safety Analysis. Norbert Helle has more than 15 years experience working in food safety and environmental analysis for various German Federal and State agencies, mainly in the field of HPLC–MS. He chairs two working groups under the German § 64 Foodstuffs and Commodities Act charged with developing food safety analysis methods for phycotoxins and biogenic amines. 

 is the Head of Research and Development at TeLA GmbH, including method development on HPLC–MS, HPLC–MS–MS and ICP–MS instruments for the determination of residues of pesticides, biotoxins and drugs. She oversees various national research projects on preparative and analytical HPLC–MS and HPLC–MS–MS work in the area of food safety. Meike Baden is a member of a working group under the German § 64 Foodstuffs and Commodities Act charged with developing food safety analysis methods for phytotoxins. 

Articles

This article describes the latest developments in LC-MS and GC-MS that reveal the mysteries of this ancient art.

Advances in the Ageing Chemistry of Distilled Spirits Matured in Oak Barrels

In the second instalment of this series, John Hinshaw discusses the use of gas switching valves for column backflush and two-dimensional series-bypass applications.

Multidimensional gas chromatography (MDGC) as practiced today may be divided into two main application areas: Classical MDGC and comprehensive two-dimensional gas chromatography (GC×GC).  

Classical MDGC involves redirecting a portion, or portions, of the eluent from one column to another column with different separation characteristics, at times that are synchronized with the elution of specific peaks (or groups of peaks) from the first column.  

The more recent technique of GC×GC also involves the transfer of the eluent from one column to another, but the transfer occurs cyclically and continuously across the entire chromatogram without targeting the timing of individual peaks. All of the eluent from the first column is transferred to the second column so that all peaks are separated by both columns. The comprehensive approach yields resolving powers that approach the product of the first and second columns with a corresponding leap in data quantity and complexity. Increased knowledge and skills are required from the analyst to use this approach effectively in the routine laboratory.  

This month's "GC Connections" will describe some basic examples of classical MDGC and offers insights into the more advanced methods of GC×GC. 

The injection of gaseous samples from a fixed-volume loop or the swapping of detectors during a GC analysis are two simple examples of ways that chromatographers use gas switching valves (GSVs). 

The previous instalment (1) showed how these fundamental valve techniques work for packed and micropacked GC columns. Chromatographers also use valves in more complex ways, such as reversing the flow of carrier gas through a column, or to manipulate and redirect gas streams between two or more columns in multidimensional GC.  

MDGC techniques enable the separation of a wide range of components that might not be achievable with a single column or in a single analysis. Separating a wide range of compounds with conventional single-column GC may require column temperature programming across a relatively large temperature range while MDGC could provide an equivalent separation isothermally, sometimes in combination with carrier gas pressure programming. 

Dual-parallel-column analyses are possible, too, but sometimes it is not desirable to inject the entire sample onto both columns, especially if all injected substances cannot be cleared from both columns before a subsequent injection.  

Performing a pair of discrete separations one after the other is also possible, but this method would also place the entire sample into each column, as well as taking the sum of both analysis times to be completed.  

Before looking at multidimensional applications, we'll first take a look at a simple column flow-reversal application. 

Chromatographers use column backflushing when the sample contains residual materials that would not otherwise be eluted from a column until well after all peaks of interest have been eluted; as an alternative to high-temperature column oven excursions where the upper temperature limit may be curtailed by the maximum temperatures allowed for the valves or column; or if temperature cycling is not otherwise required. Column and valve lifetimes can be extended by avoiding repeated high temperatures over long time periods. Backflushing works by removing late-eluting impurities from a column that might otherwise interfere with subsequent analyses. 

Backflushing of a single column can be accomplished using the 6-port valve arrangement shown in Figure 1. In Figure 1(a) the 6-port valve starts in the backflush or standby position: carrier gas flow proceeds through the column from left to right as shown. The set-up is ready for injection after the sample fills the sample loop.  

 When the valve is moved to position (b) the sample loop is flushed by carrier gas flow onto the right-hand column connection as the column flow direction reverses and starts going from right to left. As long as the sample loop pressure is low, there is no danger of the sample flowing backwards into the carrier gas pneumatics because the carrier gas pressure is higher than the low-pressure end of the column at valve port 2 at the moment of injection. 

The valve remains in the (b) position until the last peak of interest has been eluted. The valve is then returned to position (a), which causes the column flow to be restored to its original left-to-right direction (backwards in relation to the direction that the carrier gas flowed during peak elution). Any solutes or impurities left in the column at the backflush time will begin to move in reverse, towards the detector. As the total run time approaches twice the backflush time — assuming the carrier gas flow or pressure are not modified — residual material will be backflushed through the detector. At a time equal to approximately twice the backflush time nearly all of the material will have been flushed out. Some may remain as a result of band-broadening effects or nonlinear retention, so it is good practice not to perform another injection until sufficient additional time has passed in the backflush mode to remove all residual material. 

Figure 2(a) illustrates a backflushed chromatogram of a simple gas mixture. In Figure 2(b), the backflushed portion of the chromatogram is enlarged 200× to show a momentary baseline disturbance at the backflush time of 480 s, followed by some unidentified C

 and heavier compounds and a return to baseline before 960 s, at twice the backflushing time. Although the impurity levels are relatively low in this example, if they are not purged from the column after each run such extraneous non-eluted material could build up over many runs and cause baseline instability. Other examples include natural gas analysis with grouping of C

+ compounds, environmental extracts with high-boiling matrix compounds and food and beverage volatiles.  

The example shown here was performed on a micropacked column. The principles of backflushing for open-tubular columns are similar, but because of the low dead-volume requirements such separations may be better managed with various fluidic capillary switching devices. A future instalment  of "GC Connections" will address these configurations and their applications.  

Some columns yield good separations for certain compounds but not for others. For example, a molecular sieves column discriminates between various light gases such as hydrogen, oxygen, nitrogen and carbon monoxide, but fails to separate water or carbon dioxide.  

These two compounds will contaminate a molecular sieves column until baked off at high temperatures, and if they are allowed to accumulate they will degrade the separation of other compounds on the molecular sieves column. On the other hand, certain porous polymer columns easily separate water and carbon dioxide as well as a variety of hydrocarbons and other gases, but may not produce acceptable separations of those gases handled well by molecular sieves.  

It is possible to combine the separation capabilities of these two columns for a complete separation of fixed gases and C

 hydrocarbons. The rest of this instalment will address the separation of these solutes in a couple of different ways. Other examples for different sets of solutes are easy to find in the chromatography literature. 

Two popular methods for this type of combined column two-dimensional separation are sequence-reversal and series-bypass operation. In both cases, the flow and connections between two columns are manipulated during the analysis via valve operations so that each column receives and separates the peaks that it is best suited for. 

The main differences between sequence reversal and series-bypass operations is that the former uses a single detector or two detectors in series, while the latter can use two discrete detectors for true dual simultaneous separations. Column sequence reversal is illustrated by the flow diagram in Figure 3; this configuration uses a 10-port valve. The details are not shown here for reasons of space, but several examples can be found on the manufacturers' websites.  

In sequence reversal some care must be taken to ensure that peaks from one column are not co-eluted with peaks from the other. Figure 3(a) illustrates the column flows in the first step of injection with sequential flow from column 1 to column 2. 

The sample is injected from the sample loop into column 1. The initial, partially-separated peaks from column 1 enter column 2 until those peaks destined to be separated on column 2 have been transferred. The valve is then moved to change to flow as shown in Figure 3(b); the sequence of the columns is reversed.  

Peaks remaining in column 1 continue to partition and move along the column until they are subsequently eluted and detected. Simultaneously, peaks from column 2 are eluted into column 1, pass through it with little to no additional retention, and enter the detector. Co-elution of peaks in a complex mixture such as in this example may be difficult to avoid. In some applications it may be necessary to add a second valve to stop flow in column 2 and delay its peaks from entering column 1 until the existing peaks in column 1 have been eluted. 

A related configuration keeps the second column flowing to the detector, but backflushes the first column to vent after the solutes to be separated on the second column have left the first column. An independent carrier gas source may be used for backflushing the first column. This configuration is sometimes called pre-column backflushing, and can also be found in manufacturers' literature. It is useful when only the peaks from the second column are to be analysed, and the remaining peaks in the first column can be discarded. 

For the application at hand, what we really want to accomplish is the simultaneous determination of all of the peaks. This can be made possible by combining the two applications discussed so far, backflushing and column-to-column flow redirection. The backflush application shown in Figure 1 is augmented by replacing the 6-port valve with a 10-port valve, and an additional 6-port valve, carrier gas source and detector are added. The 10-port valve provides the backflush functionality, and the 6-port valve controls the destination of the effluent from column 1 — either to column 2 or to detector 1 through a restrictor — while generating functionality similar to that of the sequence-reversal configuration of Figure 3. 

Figure 4 illustrates the flow paths of a series-bypass backflush configuration. In Figure 4(a) residual material from a previous injection is backflushed from column 1 to detector 1. The configuration remains in this position until the sample loop is filled and injection may be started. Figure 4(b) shows the flows for injection and series operation of column 1 into column 2. Here, the contents of the sample loop are injected into column 1 and the initial unresolved fixed-gas peaks plus methane are passed in series onto column 2. After the methane peak passes into column 2, the valves are repositioned to produce the flow paths shown in Figure 4(c).  

Column 1 now bypasses column 2, and both columns flow to their respective detectors. The solutes that were initially passed from column 1 to column 2 are eluted into detector 2, while the remaining target peaks from column 1 are eluted into detector 1. After the column 1 target peaks have been eluted, the flows are reset to position (a), so that any remaining contaminants in column 1 are backflushed to detector 1. The restrictor, which has similar flow characteristics to column 2, serves to maintain a more constant flow in column 1 as it is switched in front of column 2 and back again.  

The valve plumbing for the series-bypass backflush configuration is shown in Figure 5. The 10-port valve provides sample injection and backflush functionality, while the 6-port valve selects whether column 1 effluent is directed to column 2 or to detector 1 via the restrictor. The valve positions in Figure 5 that correlate with the three flow configurations of Figure 4 are listed in Table 1.  

Table 1: Valve positions for Figure 5 to achieve flow paths diagrammed in Figure 4. 

The timing of the bypass operation is important. If the first column is bypassed around the second column too soon, then some of the peaks that should enter the second column do not. If the bypass time is set too long, then some of the peaks that should remain in the first column can enter the second column instead. Figure 6 shows the effects of increasing the bypass time by 20s. The methane peak is split between the first and second columns with a 60-s bypass time. Increasing the bypass time by 20 s results in all of the methane appearing on the second column. This timing must be revalidated any time the flow, temperatures or the columns themselves are changed. The timing of the backflush operation, at 480 s in this example, is not as critical and can occur at any time that the transient spike does not interfere with a peak, once all of the peaks of interest have been eluted from column 1.  

This month's instalment examined the basics of valve applications for backflushing, sequence reversal and pre-column bypassing. We then combined these basic techniques to create a hybrid dual-valve configuration capable of separating both fixed gases and hydrocarbons in a single two-dimensional analysis. MDGC presents a powerful suite of techniques that analytical chemists can use to tease apart solutes that might otherwise be inseparable, at least without significant additional effort. The next instalment in this series will discuss multidimensional separations as applied to open-tubular columns; including fluidic flow switching and applications to heart-cutting, selectivity tuning of serially connected columns and GC×GC separations. 

"GC Connections" editor John V. Hinshaw is senior research scientist at BPL Global Ltd, Hillsboro, Oregon, USA and a member of 

 editorial advisory board. Direct correspondence about this column should go to 

 editor, Alasdair Matheson, 4A Bridgegate Pavilion, Chester Business Park, Wrexham Road, Chester, CH4 9QH UK, or e-mail: 

GC Connections

Discussion about the use of gas switching valves for column backflush and two-dimensional series bypass applications.

Valves for Chromatography (Part 2): Applications

The method developed runs on a fully automated platform integrating solid-phase extraction (SPE) and LC (Symbiosis Pico) without any manual pretreatment step(s). Automatic pretreatment results in high throughput, low CVs and high precision. Sample preparation has been simplified to just one operator step before the system takes over the automated analysis of the samples.  

A Spark Holland Symbiosis PICO system with a Mistral column oven and an Applied Biosystems/MDS Analytical Technologies API 4000 LC–MS–MS system are used for this application. The API 4000 is equipped with a Heated Nebulizer (APCI) probe. 

 is Spark Holland's unique (u)HPLC solution with integrated on-line SPE-LC–MS automation. The system offers large flexibility in processing different types of samples selecting one of the three fully automated operational modes LC; XLC (extraction liquid chromatography) and mXLC (multidimensional on-line SPE) 

Sample preparation consists of diluting a human sample with a special protein disrupting buffer, followed by mixing. Samples are then placed in the Symbiosis autosampler. 100 μL of sample is injected on a Spark HySphere C8EC-SE sorbent SPE cartridge for on-line sample clean-up and concentration.  

The mobile phase consists of formic acid, water and methanol with separation on a Merck KGaA Chromolith SpeedROD RP-18e (endcapped), 50 × 4.6 mm HPLC column. An example chromatogram of the separation achieved is shown in Figure 1.  

A calibration curve was created by injecting a full set of calibration standards ranging from 5 to 500 ng/mL spiked human plasma. The calibration curve is calculated with a R

  of 0.999 with a 1/X weighting for both compounds.  

Table 1: Accuracy and precision of the QC samples. 

This method was evaluated using spiked human sodium heparin plasma samples. The developed on-line SPE-MS–MS method has an absolute recovery of more then 90%, and calibration curve with a linear range from 5–500 ng/mL (R = 0.999 for both compounds). The developed method has an accuracy ranging from 89–117% and precision of 0.1–7.9 %CV. The estimated quantification limits for both analytes are more than sufficient to allow the analytical method to be used for therapeutic drug monitoring. When injecting 100 μL, containing 50 μL of plasma, a detection limit of 5 ng/mL is achieved. 

2. Applied Biosystem Vitamin D i-method (User Contributed i-method by Spark). 

For more information please send an e-mail to:  

 P. de Keyserstraat 8, PO box 3887800 AJ Emmen, The Netherlands 

Departments

Determination of 25-mono-hydroxy-vitamin D2 and D3 in Human Plasma Using a Symbiosis Pico System: ACN Free On-line SPE-LC–MS–MS

Experimental design or sequential approaches can be used to optimize separations in pharmaceutical analysis However, the choices made by the analysts setting up the experiments are not always the best and may create problems. In this article the different steps of an experimental design-based approach are considered and the problems that arise from the wrong choices are discussed.

In previous articles we have discussed the experimental designs that can be applied in method development and optimization as well as the analysis of this data (1–4). In recent articles we described a method development approach for drug impurity profiles (5,6). This approach is a variant on the common experimental design one. In an experimental design approach one first determines the most important factors from screening design results. The responses of the method are then modelled as a function of these latter factors and the optimal factor combination determined from the results of response surface designs. For drug impurity profiling we could skip the screening step since we usually know in reversed-phase chromatography which factors have the largest influence on the selectivity. This allows the use of a sequential approach in method optimization, where some steps included small designs (5,6). 

However, in some manuscripts originating from both the pharmaceutical industry and the academic world, some steps from the experimental design [or design of experiments (DoE)] approach are not properly applied, leading to problems: either with the applied designs, in the practical execution of the design or in the data handling steps. We will limit ourselves in the rest of this article to the application response surface designs. The different steps in a DoE approach are the selection of the factors and their levels, the selection of the appropriate design to examine the factors, the selection of the response(s) to consider, the execution of the design experiments and the data handling. 

The number of factors to optimize should preferably be low (i.e., two or three). These most important factors are, for example, selected based on the results of a screening design performed earlier. Regularly, optimization designs are executed for four to six factors. Such designs require a considerable number of experiments, but determining the best conditions from it is far from obvious and we will discuss this later. Moreover, it is rare that such a high number of factors would significantly affect the method. Therefore it is better to make an 

 selection of the important factors in a preceding screening step. The factors are best defined in such a way that they have a physico-chemical meaning that is easy to understand by the users. For example, pH can be defined as a factor. Everybody knows its meaning and its influence could be derived and interpreted from the design results. Even though pH could also be split in the concentrations of its acidic and basic compounds that might be included as two factors in a design, the physicochemical interpretation of their effects is far less evident. 

Factor levels during optimization are selected in an interval that is as broad as possible and that are, for example, limited by the product to be optimized or the limits of the applied technique. For instance, if DoE is applied to optimize a formulation, such as a tablet, then the most extreme levels (as well as their combination) should be tested preliminarily to verify whether these extremes still result in the desired product (i.e., the tablet). In chromatography, the pH, for example, might be limited depending on the type of columns used. 

Based on the number of factors to be examined, an experimental design is selected. We discussed the most frequently applied response surface designs in reference 3. It concerns designs with at least three levels for each factor. When examining four or more factors the required number of experiments is high. For example, when applying a central composite design studying 4, 5 or 6 factors requires at least 25, 43 and 77 experiments, respectively. This is another reason to keep the number of factors examined limited. Often a number of replicated experiments, usually of the centre point, are added to the experimental set-up. They are used to estimate the experimental error and are later applied in the data interpretation to indicate non-significant terms in the models built. This will be discussed later. 

The responses measured should allow the prediction of the best conditions, such as those where the worst separated peak pair is best separated. At first sight a response indicating the degree or the quality of separation, such as the selectivity factor, α, or the resolution, Rs, seems appropriate for this purpose. To predict the best conditions the response is modelled as a function of the factors examined, usually applying a quadratic model. For instance, for considering two factors:  

 the model coefficients. For more than two factors similar models can be built containing more terms (4). 

However, modelling resolution or the selectivity factor from the measured design values is usually not recommended, unless no changes in elution sequence occur among all the experiments, which in method optimization is very improbable. When changes in elution sequence occur, there are then two objections against the modelling of the above responses. 

First, a situation where two peaks A and B are equally separated but eluting in a different sequence, that is once as AB and once as BA, both are indicated by the same number to represent a different situation. In the factor's domain between both conditions the separation will go through a situation of co-elution or Rs = 0. When modelling the above, the different situations AB and BA are not distinguished and no model will pass through zero in the intermediate domain. Thus to distinguish between both situations one should give one of the two resolutions a negative sign [e.g., Rs(AB) = 1.5 and Rs(BA) = –1.5]. In a study where resolution was modelled, it was reported that "almost all models for Rs had serious lack of fit problems". The reason for these problems can be easily found in the above. The optimum was then selected as the best result from the design experiments. There is nothing wrong with that approach, but then one does not use all the available information. 

Second, only the resolution between consecutive peaks is of interest to find the best conditions. However, the problem is that because elution sequence changes occur between design experiments, one does not know which peak pairs are relevant at a given set of conditions. Consequently, one should model the resolution between all possible peak pairs. This means 15 models for a mixture with six compounds and 78 models for a mixture of 13 compounds. Such modelling becomes tedious, while a majority of models will in fact be redundant. For the two reasons above one does better not to model resolution (or selectivity factor) during separation optimization. 

Alternatively, the retention times (and occasionally the peak widths) will be modelled. This approach requires one (occasionally two) models for each compound. Peak widths are also modelled for isocratic separations, while for gradient elution one observes from the measured peak widths whether they can be considered constant or not and thus don't need to be modelled. Then the relevant resolutions are calculated from the modelled responses for given experimental conditions, because;  

 are the retention times and baseline peak widths of two adjacent peaks. 

In a DoE approach a quadratic model (Equation 1) is usually built. In a number of publications the results of replicated experiments (e.g., at centre point level) are used to estimate experimental error and to determine which coefficients of the model are non-significant (i.e., equal to zero). Very often nothing is done with the observed results. Normally a new model without the non-significant terms should then be build. However, when one uses either the first model or the second to select the optimal conditions, it will not make a large difference. The models determine regions with proper response characteristics. When considering the region determined by the full quadratic model, one does not expect the selection to be largely changed from a model modelling the same data but where the non-significant contributions to the model are eliminated. 

Often in method optimization the appropriate conditions are visually derived from the response surfaces or the equivalent contour plots (Figure 1). These are visualizations of the quadratic model that has been built. Now when examining three factors, one has to fix one factor at a given value to be able to visualize a part of the entire response surface. This is similar to selecting a chromatogram from an LC–DAD measurement.  

Figure 1: Response surface and its corresponding contour plot, representing the behavior of a response as a function of two factors. 

The chromatogram selected at low wavelength does not necessarily resemble that at an intermediate or at a high wavelength. Each of them only shows a tiny part of the entire data set. The response surfaces observed at low, intermediate and high fixed value of the third factor also do not necessarily resemble each other and they also each show a fraction of the entire response surface. When evaluating four, five or six factors in an optimization step, one has to fix two, three or four factors to be able to draw a response surface for the two remaining. The more factors that need to be fixed the smaller the fraction of the entire response surface one is visualizing and observing. In fact one does not have an idea whether the observed best conditions represent a local optimum, which can be far away from the global one, or not. 

Now when considering our specific problem of separating a mixture of compounds, we will not determine the optimum from the response surface visualization because only models for retention time (and occasionally for peak width) are available. To find the conditions where the worst separated peak pair is separated best, a grid over the entire domain is considered. This approach is explained in detail in references 5 and 6. For each factor in its examined interval a feasible step size is defined. For example, for percentage of modifier  the step size might be 1%, 2% or 5%; for the pH it could be 0.1 or 0.2 pH units. These steps determine a number of feasible levels within the factor range, for example 20 for %modifier and 30 for the pH. The above grid is composed of all combinations of the different factor levels. For example, for pH and modifier in the above example, the grid contains 600 points (i.e., 20 × 30). When four or more factors are considered, easily ten- or hundred thousands of grid points are obtained, which makes the calculation work tedious. Occasionally then broader steps can be considered resulting in fewer grid points. 

At each grid point one predicts the retention time of a given compound from the model built and considers its corresponding width. The retention times (with their widths) are then sorted from high to low (representing consecutive peaks) and the resolutions between consecutive peaks are calculated. By this approach only the relevant resolutions are considered. At each point the smallest of these resolutions is selected because it represents the worst separated peak pair at that set of conditions. From all minimal resolutions one can derive the different conditions where the worst separated pair is best separated or where it is baseline separated, Rs > 1.5. This approach leads to the indication of the conditions were best separations are predicted. It does not require any visualization but indicates the proper optimum, even for higher numbers of examined factors, though there the calculation load is high. The predicted separation at the selected set of conditions is then verified experimentally. 

The following recommendation can be given from a practical point of view to separate mixtures with a considerable number of compounds (e.g., in drug impurity profiles), and regards the gradient or solvent strength conditions: When measuring experimental design results to apply an approach as described above, take care to choose the initial gradient or solvent strength conditions such that the elution time differences between the first and last eluting peaks are large enough. If not, one risks performing lots of calculations and ending up with best conditions where not all peaks are baseline separated. Speeding up the analysis once baseline separation is obtained belongs to the fine tuning of the method and can be done rather quickly in different ways (e.g., gradient adaption, flow-rate changes, column dimension changes), while a repetition of the experimental design conditions at an adapted gradient is much more tedious. 

 is a professor at the Vrije Universiteit Brussel, Belgium, department of Analytical Chemistry and Pharmaceutical Technology, and heads a research group on chemometrics and separation science. 

5. B. Dejaegher, M. Dumarey and Y. Vander Heyden, 

6. B. Dejaegher, M. Dumarey and Y. Vander Heyden, 

Practical Data Handling

The various steps of an experimental design-based approach are considered and the problems that arise from wrong choices are described.

Common Problems with Method Optimization in Pharmaceutical Analysis

A robust, selective and sensitive separation method for the determination of four steviol glycosides in stevia plant matrix is demonstrated. In addition to the two calibrated substances two more glycosides could be found in stevia plant extracts by LC–MS. Using a popular Eurospher NH

 phase and applying moderate separation conditions results in an inexpensive and precise analysis method.

Steviol glycosides are responsible for the sweet taste of stevia plant leaves. These compounds are 75 to 450 times sweeter than sucrose without the typical bitter aftertaste of some well-known artificial sweeteners. In the course of the accreditation of these four steviol glycosides by the public health authority a stable and sensitive analytical quantitative separation method is highly recommendable.  

The separation was performed in isocratic mode on a KNAUER Smartline HPLC system with low pressure gradient capability, degasser, autosampler and column oven.  

The separation could be achieved in 10 minutes. The limit of detection (LOD, S/N = 3) was determined for stevioside at 7.5 ng and for rebaudioside A at 16.0 ng.  

A fast and very good separation of four steviol glycosides in Stevia leaf extract samples with excellent peak symmetry is easily accomplished by reversed-phase HPLC using the Eurospher NH

 column in combination with a Smartline HPLC system. The simple isocratic method with moderate parameters guarantees robust and sensitive results over a long time period. The method allows safe quantification. Contact 

 for a more detailed version of this application.  

tel. +49 30 809727 0  fax +49 30 801501 0  

Separation of Steviol Glycosides

In the first two instalments of this series, we've looked at different factors that influence selectivity in reversed-phase liquid chromatography (LC) separations. First, we saw some of the benefits of changing solvent type,

  such as a change from methanol to acetonitrile or tetrahydrofuran. Next, we looked at how a change in the solvent strength

  — the percentage of organic solvent in the mobile phase — was an easier adjustment to make and often had suitable "leverage" to pull apart problem peak pairs. This month's "LC Troubleshooting" will focus on another common technique used to change peak spacing in a chromatogram: changing from one column type to another. 

When I first started using LC, it was thought widely that all C18 columns would give equivalent separations. In fact, the 

(USP) in its column classification scheme classifies C18 columns as "L1 — octadecyl silane chemically bonded to porous silica or ceramic micro-particles, 1.5 to 10 µm in diameter, or a monolithic silica rod".

  Over the past 40 years, however, we've all come to recognize that all C18s are not created equal — many will give dramatically different separations. In the current listing, the L-classification scheme contains 74 different column types,

  including C18, C8, phenyl, cyano and many others. Even columns that have the same label, such as C18 or phenyl, may have significantly different selectivity characteristics. It is not clear how many different reversedphase columns are available, but by some estimates there are at least 1000.

  In our work at LC Resources, we've tested more than 500 different reversed-phase columns. 

Figure 1: Comparison of simulated chromatograms for C18 columns from three manufacturers. Column dimensions: 150 mm Ã 4.6 mm, 5 Î¼m particles; mobile phase: 50:50 acetonitrileâ30 mM phosphate buffer (pH 2.8); temperature: 35 Â°C; column efficiency N: 10000. Retention normalized to peak 1. See Table 1 for column information and text for details. Peaks: 1 = anisole, 2 = n-butylbenzoic acid, 3 = toluene, 4 = mefenamic acid, 5 = ethylbenzene, 6 = trans-chalcone. 

So how different are C18 columns from each other? To illustrate this, I've chosen a set of six test compounds that should be fairly well-behaved (see caption of Figure 1 for their identities). That is, at low pH (pH 2.8), they are neutral or un-ionized, so ionic interactions with the column should be minimal. I arbitrarily chose three C18 columns from our database.

  These are from well-known sources and are all type-B, high-purity silica columns. These are simulated separations on 150 mm × 4.6 mm columns packed with 5 µm particles and generating 10000 plates. For simplified comparisons, I've normalized retention to the first peak (anisole). The first column [Figure 1(a)] is our reference column. You can see that peaks 4 and 6 overlap and the resolution is marginal for peaks 2 and 3. If we change to another manufacturer's C18 column [Figure 1(b)], the separation of peaks 2 and 3 is improved, and the separation of the last three peaks has changed, but it is no better than the reference column. A third manufacturer's C18 [Figure 1(c)] is no better — it shows a good separation for peaks 2 and 3, but peaks 4 and 6 still overlap. You can see in Figure 1 that, although the first three peaks are separated by all three columns, the selectivity (relative peak spacing) is different for these peaks. And none of the three columns can separate the last three peaks — even the peak order changes! So you can't arbitrarily swap one C18 for another and expect the same results. But on the other hand, the differences aren't enough that we can make the same change and expect to solve a separation problem. Although there is likely at least one manufacturer's C18 column that will separate all six compounds, changing from one to another in an effort to get the desired separation is not a very good strategy. 

Another technique that has been around as long as I have been doing chromatography is to change the stationary phase. That is, users change from a C18 to a phenyl or cyano or embedded polar phase. We can use the same sample and conditions to see what happens in this case. I've chosen columns from the same manufacturer as the reference column of Figure 1(a). This separation is repeated in Figure 2(a) (with a slightly extended time axis). The sample was then run on the same brand of column from the same manufacturer, but with a phenyl stationary phase for the separation of Figure 2(b). The separation of the phenyl column might be expected to be much different from the C18, because there is aromatic nature to the sample components. The phenyl column separates all the peaks, but not to baseline. We could probably adjust the mobile phase or reduce the particle size (or both) to get baseline separation with this column. Figure 2(c) shows the separation on the same manufacturer's embeddedpolar-group (EPG) phase. This column provides a good separation of all the peaks, certainly the best so far. Notice also that peaks 2 and 3 are reversed with the EPG column. The differences between the columns are assumed to be only the stationary phase — the silica particle chemistry should be the same. This technique, of changing to a different stationary phase from the same manufacturer, is one that many people have had success with over the years, but it still involves trial and error, without a certainty of a satisfactory separation. 

Figure 2: Comparison of chromatograms for (a) C18, (b) phenyl and (c) embedded-polar-group columns from same manufacturer. Other conditions as in Figure 1; see text for details. 

Still another way to pick a column that is likely to have different selectivity is to use the column database on the USP website.

 and elsewhere (for example, see reference 8), this database with more than 500 reversed-phase columns can be used to pick similar (equivalent) or different (orthogonal) columns compared to some reference column, such as the one you have already tried. By choosing the "view different columns" option, you can see what columns are most different from the reference column. I tried this out for the present sample containing some acids and neutral compounds at low mobile-phase pH. As my reference column, I chose the same C18 column we used in Figures 1(a) and 2(a), and have repeated the separation as Figure 3(a) (different time axis than the other figures). One of the most orthogonal columns with type-B silica particles (use of the older, low purity, type-A particles is discouraged for new separations) is the one used to obtain the separation shown in Figure 3(b). This is an EPG column from a different manufacturer than the reference column of Figure 3(a). So we've taken advantage of both a change in the stationary phase and a change in the silica source for added selectivity potential. The change is dramatic. Peak 5 has moved forward between 1 and 2, and peak 6 has moved up between 2 and 3. This is definitely a change in selectivity. There is no guarantee that the recommended column will separate all the compounds in your sample, but there is a high probability that the separation will look much different, especially if you take advantage of other selectivity variables. In one collaborative study,

  several pharmaceutical laboratories tested this approach on typical samples that they encounter, and in all 11 of the cases checked, the predicted orthogonal column indeed gave a significantly different separation. 

Figure 3: Comparison of separations obtained using (a) a C18 column from one manufacturer with (b) an embedded-polar-group column from a second manufacturer. Other conditions as in Figure 1; see text for details.  

We can make visual comparisons and see that, for the present sample, a change from the column of Figure 1(a) [same as Figures 2(a) and 3(a)] to the column of Figure 2(c) or 3(b) would allow us to separate all the peaks. But which of these separations gives us the largest change? When comparing the selectivity changes when a change in column is made, it is nice to have a quantitative measure of the change. One way to do this is to make a graphic comparison of selectivity. This can be achieved by plotting the retention on one column vs the retention on a second column. An example of this for the columns of Figure 1(a) and 1(b) is shown in Figure 4(a). (Note that plots like this are most useful if we plot the log of the retention factor, 

, for each column.) The scatter about the trend line in the plot is a measure of how different the selectivity is between the two columns. The data cluster tightly about the trend line, and 

 = 0.97 (Table 1) confirms this. Contrast this result to that of Figure 4(b), where the columns of Figure 3(a) and 3(b) are compared. Here, there is much more scatter (

 = 0.49), showing that the selectivity is very different between the two columns. 

Figure 4: Plots of log k vs log k for sample of Figure 1. (a) Column of Figure 1(a) vs 1(b); (b) column of Figure 3(a) vs 3(b). Other conditions as in Figure 1; see text for details. 

An actual calculation of the change in selectivity, α, is an even better way to compare two columns. Recall that  

                                       [1] 

 are the retention factors for two peaks. The change in log α between the two columns is related to how far the data points are from the trendline in plots of log 

, as in Figure 4. The larger the average change in log ͍, |δlog α|avg, between two columns, the more different the chromatogram will appear. This measure of change in selectivity, |δlog α|avg, is directly related to the standard error (SE) of the plot as 

|δlog α|avg = 1.4 SE                      [2] 

so it is easy to calculate from plots such as Figure 4. Other work has shown that when selectivity is changed in a reversed-phase separation, whether by a change in column or some other variable (for example, solvent type or solvent strength), a value of |δlog α|avg ≥ 0.1 gives a high probability that the separation will be orthogonal (or significantly different).

  For the present examples, I have made these calculations for the various columns when compared to the reference column [Figures 1(a), 2(a), 3(a)] and summarized them in Table 1. You can see that the column of Figure 2(c) almost reaches the desired change, but that the column of Figure 3(b) has a value of |δlog α|avg that is more than twice as large as the minimum recommendation. This confirms once more our visual observation of the separation that is most different from the original. 

Table 1: Comparison of column selectivity. 

  gives still another measure of differences in column selectivity. This is the 

-value calculated from the column characteristics in the database.

  In one study with 67 different compounds of widely varying structure, including acids, bases and neutrals, it was determined that 

  This means that the database can be used to look for columns with large 

-values compared to the current (reference) column, thus giving you confidence that peak-spacing changes will be significant. I have tabulated the 

-values in Table 1 for the six columns we have been examining and the selected sample (neutrals and un-ionized acids). You can see that the column of Figure 3(b) has 

 = 55, whereas all of the other columns have 

 ≤ 16, once more confirming the visual observations that this is the most orthogonal column of the set. 

We have seen that a change in column chemistry can be a powerful tool to change selectivity in a reversed-phase separation. A change from one manufacturer's C18 to one from another manufacturer generally gives marginal gains, as was seen in Figure 1. A more effective way to make the change is to try a different stationary phase. If this change is made within one product line of columns, where the same silica support material is assumed, the change will result from a change in the bonded phase alone. This was shown in Figure 2 for three different bonded phases from the same product line from a single manufacturer. To gain even more leverage in the separation, a change in the base silica along with a change in bonded phase can be made. Although this may be possible within one manufacturer's products, such as by changing to a different bonded phase from a different product line of columns, such a change is more likely if the change is made in both the manufacturer and bonded phase, as shown in Figure 3.  

Using a change in the column to change peak spacing comes with risks of making a poor choice [as in Figure 1 or Figure 2(b)]. Such results also can be costly, because columns are expensive and research time is valuable. A more productive technique is to use some foreknowledge of selectivity differences between specific columns, such as that available in the USP database.

  By using the database, it is easy to identify columns that are most likely to have different selectivity.  

When exploring changes in selectivity, it probably will be more practical to explore solvent-type selectivity

  before column-type selectivity, just because it is less expensive. However, armed with the information we've covered in this discussion, you can make intelligent choices about a change in column, so you can increase the return on the investment in a new column in an effort to change selectivity. When this is taken into account, you may find that screening several columns of different selectivity is a fruitful investment of time early in the separation development process. 

"LC Troubleshooting" editor John Dolan is vice president of LC Resources, Walnut Creek, California, USA. He is also a member of 

 Europe's editorial advisory board. Direct correspondence about this column should go to to 

 Europe editor, Alasdair Matheson, at Advanstar Communications, Bridgegate Pavillion, Chester Business Park, Wrexham Road, Chester CH4 9QH, UK, or e-mail 

8. L.R. Snyder, J.W. Dolan and P.W. Carr, 

LC Troubleshooting

How to choose another column to try and solve a separation problem.

LCGC Asia Pacific-09-01-2011

Advances in the Ageing Chemistry of Distilled Spirits Matured in Oak Barrels

Valves for Chromatography (Part 2): Applications

Determination of 25-mono-hydroxy-vitamin D2 and D3 in Human Plasma Using a Symbiosis Pico System: ACN Free On-line SPE-LC–MS–MS

Common Problems with Method Optimization in Pharmaceutical Analysis

Separation of Steviol Glycosides

Selectivity in Reversed-Phase Separations (Part 3): Column-type Selectivity